Ciência de Dados para Negócios: Big Data for Finance Project

Resumo

teste de futuro para as ações

Intro

escrever

Code

library(tidyverse)
library(tidymodels)
library(modeltime)
library(timetk)
library(purrr)
library(tidyquant)
library(tsibble)
library(prophet)
library(feasts)
library(fable)
library(fabletools)
library(lubridate)
library(tictoc)

Carregamos os dados:

Code

tickers <- c(
         "BRFS3.SA",
  "JBSS3.SA",
  "BEEF3.SA",
  "MRFG3.SA",
  "TSN",
  "HRL",
  "GIS"
)

Então baixo os dados via Yahoo!Finance:

Code

portfolioPrices <- NULL
  for ( Ticker in tickers )
    portfolioPrices <- cbind(
      portfolioPrices, 
      quantmod::getSymbols.yahoo(
        Ticker,
        from = "2019-01-01",
        auto.assign = FALSE
      )[,4]
    )

portfolioPrices <- portfolioPrices[apply(portfolioPrices, 1, function(x) all(!is.na(x))),]

colnames(portfolioPrices) <- c(
  "BRFS3.SA",
  "JBSS3.SA",
  "BEEF3.SA",
  "MRFG3.SA",
  "TSN",
  "HRL",
  "GIS"
)

# Visualizar com DT
#DT::datatable(tail(portfolioPrices), options = list(pageLength = 10, scrollX = TRUE))

Visualizando os dados dos nossos últimos retornos dos preços, temos:

Code

log_returns <- log(portfolioPrices) - log(lag(portfolioPrices))
log_returns <- na.omit(log_returns)
log_returns <- log_returns |> 
  timetk::tk_tbl(preserve_index = TRUE, rename_index = "date")

tail(log_returns)

Code

ln_returns <- log_returns

ln_returns |> as.data.frame() |>
  dplyr::mutate(
    time = seq_along( TSN )
  ) |> select(-date) |>
  tidyr::pivot_longer(
    !time,
    names_to = "Variables",
    values_to = "Value"  
      ) |>
  dplyr::group_by(Variables) |>
  timetk::plot_time_series(
    time,
    Value,
    .interactive = F, # Change for TRUE for better visualization
    .facet_ncol = 2,
    .smooth = FALSE
  ) +
  ggplot2::theme(
    strip.background = ggplot2::element_rect(fill = "white", colour = "white")
  )

Modelagem com fpp3 e validação cruzada temporal

Precisaremos fazer um forecasting de curto prazo com nossos dados históricos de retornos pra formularmos nossas recomendações posteriores de compra, venda e espera:

Vamos começar com uma série por vez $\Rightarrow$ TSN

Code

# Primeiro converto pra tsibble

lnretTSN <- log_returns |> 
  select(date, TSN) |> 
  as_tsibble(index = date)

glimpse(lnretTSN)

Rows: 1,519
Columns: 2
$ date <date> 2019-01-03, 2019-01-04, 2019-01-07, 2019-01-08, 2019-01-09, 2019…
$ TSN  <dbl> 0.0211432803, 0.0122208208, 0.0160060857, 0.0264099860, -0.017528…

Code

treino <- lnretTSN |>
  filter_index(~"2025-01-01")

War models

Code

tic()

Modelos <- treino |>
  model(
    AjusteExp = ETS(TSN ~ error("A") + trend("N") + season("N")), # Ajuste Exponencial com auto
    
    AjExp_aditivo = ETS(TSN ~ error("A") + trend("A") + season("A")), # Ajuste Exponencial Aditivo
    
    AjExp_multiplicativo = ETS(TSN ~ error("M") + trend("A") + season("M")), # Ajuste Exponencial Multiplicativo
    
    Croston = CROSTON(TSN), # Modelo Croston
    
    HoltWinters = ETS(TSN ~ error("M") + trend("Ad") + season("M")), # Holt Winters
    
    Holt = ETS(TSN ~ error("A") + trend("A") + season("N")), # Holt
    
    HoltAmort = ETS(TSN ~ error("A") + trend("Ad", phi = 0.9) + season("N")), # Holt Amortecida
    
    Regr_Comp = TSLM(TSN ~ trend() + season()), # Regressao com tendencia e sazonalidade auto
    
    Regr_Harmonica = TSLM(TSN ~ trend() + fourier(K = 2)), # Regressao harmonica
    
    Regr_Quebras = TSLM(TSN ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais
    
    Snaive = SNAIVE(TSN), # SNAIVE
    
    Naive = NAIVE(TSN), #NAIVE
    
    Media_Movel = ARIMA(TSN ~ pdq(0,0,1)), # Media Movel Simples
    
    autoARIMA = ARIMA(TSN, stepwise = FALSE, approx = FALSE), # Auto ARIMA
    
    autoARIMA_saz = ARIMA(TSN, stepwise = FALSE, approx = FALSE, seasonal = TRUE), # AutoARIMA Sazonal
    
    #    Regr_erros_ARIMA = auto.arima(TSN, xreg = fourier(K = 3), seasonal = FALSE), # Regressao com erros ARIMA
    
    ARIMA_saz_012011 = ARIMA(TSN ~ pdq(0,1,2) + PDQ(0,1,1)), # ARIMA Sazonal ordem 012011
    
    ARIMA_saz_210011 = ARIMA(TSN ~ pdq(2,1,0) + PDQ(0,1,1)), # ARIMA Sazonal ordem 210011
    
    ARIMA_saz_0301012 = ARIMA(TSN ~ 0 + pdq(3,0,1) + PDQ(0,1,2)), # ARIMA sazonal
    
    ARIMA_quad = ARIMA(TSN ~ I(trend()^2)), # ARIMA com tendencia temporal quadratica
    
    ARIMA_determ = ARIMA(TSN ~ 1 + trend() + pdq(d = 0)), # ARIMA com tendencia deterministica
    
    ARIMA_estocastico = ARIMA(TSN ~ pdq(d = 1)), # ARIMA com tendência estocastica
    
    Regr_Harm_dinamica = ARIMA(TSN ~ fourier(K=2) + PDQ(0,0,0)), # Regressao Harmonica Dinamica
    
    Regr_Harm_Din_MultSaz = ARIMA(TSN ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = 7*30, K = 10) + fourier(period = 7*30, K = 5)), 
    
    Regr_Harm_Din_Saz = ARIMA(TSN ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = "month", K = 10) +
                                fourier(period = "year", K = 2) ), # Rgr Harm Mult Saz Complexa
    
#    Auto_Prophet = prophet(TSN), # Auto prophet
    
#    Prophet_mult = prophet(TSN ~ season(period = "month", order = 2, type = "multiplicative")),
    
#    Prophet_aditivo = prophet(TSN ~ season(period = "month", order = 2, type = "additive")),
    
#    Prophet_geom = prophet(TSN ~ growth("geometric") + season(period = "month", order = 2, type = "multiplicative")),
    
#    Prophet_memo = prophet(TSN ~ growth("geometric") + season(period = "month", order = 5) +
#                             season(period = "year", order = 2, type = "multiplicative")),
    
    Modelo_VAR = VAR(TSN, ic = "bic"), # Vetor Autoregressivo 
    
    Random_Walk = RW(TSN ~ drift()), # Random Walk com drift
    
    Rede_Neural_AR = NNETAR(TSN, bootstrap =  TRUE)#, # Rede Neural com auto AR e bootstraping nos erros
    
    #    x11 = X_13ARIMA_SEATS(TSN ~ x11()) # X11 ARIMA Seats
    
  ) |>
  
  forecast(h = "24 months") # Horizonte de projecao para os proximos 30 dias apos corte no treino

toc()

1.02 sec elapsed

Selecionamos o melhor modelo (1 fold de validação cruzada somente):

Code

Modelos |>
  accuracy(lnretTSN) |>
  arrange(RMSE) # Seleção da acuracia pelo menor RMSE para o conjunto de modelos

Gero um cenário com o modelo:

Code

fit <- lnretTSN |>
  model(
    Regr_Quebras = TSLM(TSN ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais
  )

sim <- fit |> generate(h = 30, times = 5, bootstrap = TRUE)

Plotamos os forecasts com esse modelo pra três cenários distintos no futuro:

Code

lnretTSN |>
  filter_index("2025-01-01"~.) |>
  ggplot(aes(x = date)) +
  geom_line(aes(y = TSN)) +
  geom_line(aes(y = .sim, colour = as.factor(.rep)),
    data = sim) +
  labs(title="Valores projetados de retornos de preços de contratos futuros da TSN", y="$US" ) +
  guides(colour = "none")

Code

# Primeiro converto pra tsibble

lnretGIS <- log_returns |> 
  select(date, GIS) |> 
  as_tsibble(index = date)

glimpse(lnretGIS)

Rows: 1,519
Columns: 2
$ date <date> 2019-01-03, 2019-01-04, 2019-01-07, 2019-01-08, 2019-01-09, 2019…
$ GIS  <dbl> 0.0154921374, 0.0200387411, 0.0164387227, 0.0149567729, -0.016440…

Code

treino <- lnretGIS |>
  filter_index(~"2025-01-01")

Code

tic()

Modelos <- treino |>
  model(
    AjusteExp = ETS(GIS ~ error("A") + trend("N") + season("N")), # Ajuste Exponencial com auto
    
    AjExp_aditivo = ETS(GIS ~ error("A") + trend("A") + season("A")), # Ajuste Exponencial Aditivo
    
    AjExp_multiplicativo = ETS(GIS ~ error("M") + trend("A") + season("M")), # Ajuste Exponencial Multiplicativo
    
    Croston = CROSTON(GIS), # Modelo Croston
    
    HoltWinters = ETS(GIS ~ error("M") + trend("Ad") + season("M")), # Holt Winters
    
    Holt = ETS(GIS ~ error("A") + trend("A") + season("N")), # Holt
    
    HoltAmort = ETS(GIS ~ error("A") + trend("Ad", phi = 0.9) + season("N")), # Holt Amortecida
    
    Regr_Comp = TSLM(GIS ~ trend() + season()), # Regressao com tendencia e sazonalidade auto
    
    Regr_Harmonica = TSLM(GIS ~ trend() + fourier(K = 2)), # Regressao harmonica
    
    Regr_Quebras = TSLM(GIS ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais
    
    Snaive = SNAIVE(GIS), # SNAIVE
    
    Naive = NAIVE(GIS), #NAIVE
    
    Media_Movel = ARIMA(GIS ~ pdq(0,0,1)), # Media Movel Simples
    
    autoARIMA = ARIMA(GIS, stepwise = FALSE, approx = FALSE), # Auto ARIMA
    
    autoARIMA_saz = ARIMA(GIS, stepwise = FALSE, approx = FALSE, seasonal = TRUE), # AutoARIMA Sazonal
    
    #    Regr_erros_ARIMA = auto.arima(TSN, xreg = fourier(K = 3), seasonal = FALSE), # Regressao com erros ARIMA
    
    ARIMA_saz_012011 = ARIMA(GIS ~ pdq(0,1,2) + PDQ(0,1,1)), # ARIMA Sazonal ordem 012011
    
    ARIMA_saz_210011 = ARIMA(GIS ~ pdq(2,1,0) + PDQ(0,1,1)), # ARIMA Sazonal ordem 210011
    
    ARIMA_saz_0301012 = ARIMA(GIS ~ 0 + pdq(3,0,1) + PDQ(0,1,2)), # ARIMA sazonal
    
    ARIMA_quad = ARIMA(GIS ~ I(trend()^2)), # ARIMA com tendencia temporal quadratica
    
    ARIMA_determ = ARIMA(GIS ~ 1 + trend() + pdq(d = 0)), # ARIMA com tendencia deterministica
    
    ARIMA_estocastico = ARIMA(GIS ~ pdq(d = 1)), # ARIMA com tendência estocastica
    
    Regr_Harm_dinamica = ARIMA(GIS ~ fourier(K=2) + PDQ(0,0,0)), # Regressao Harmonica Dinamica
    
    Regr_Harm_Din_MultSaz = ARIMA(GIS ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = 7*30, K = 10) + fourier(period = 7*30, K = 5)), 
    
    Regr_Harm_Din_Saz = ARIMA(GIS ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = "month", K = 10) +
                                fourier(period = "year", K = 2) ), # Rgr Harm Mult Saz Complexa
    
#    Auto_Prophet = prophet(TSN), # Auto prophet
    
#    Prophet_mult = prophet(TSN ~ season(period = "month", order = 2, type = "multiplicative")),
    
#    Prophet_aditivo = prophet(TSN ~ season(period = "month", order = 2, type = "additive")),
    
#    Prophet_geom = prophet(TSN ~ growth("geometric") + season(period = "month", order = 2, type = "multiplicative")),
    
#    Prophet_memo = prophet(TSN ~ growth("geometric") + season(period = "month", order = 5) +
#                             season(period = "year", order = 2, type = "multiplicative")),
    
    Modelo_VAR = VAR(GIS, ic = "bic"), # Vetor Autoregressivo 
    
    Random_Walk = RW(GIS ~ drift()), # Random Walk com drift
    
    Rede_Neural_AR = NNETAR(GIS, bootstrap =  TRUE)#, # Rede Neural com auto AR e bootstraping nos erros
    
    #    x11 = X_13ARIMA_SEATS(TSN ~ x11()) # X11 ARIMA Seats
    
  ) |>
  
  forecast(h = "24 months") # Horizonte de projecao para os proximos 30 dias apos corte no treino

toc()

0.99 sec elapsed

Code

fit <- lnretGIS |>
  model(
    Regr_Quebras = TSLM(GIS ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais
  )

sim <- fit |> generate(h = 30, times = 5, bootstrap = TRUE)

Code

lnretTSN |>
  filter_index("2025-01-01"~.) |>
  ggplot(aes(x = date)) +
  geom_line(aes(y = TSN)) +
  geom_line(aes(y = .sim, colour = as.factor(.rep)),
    data = sim) +
  labs(title="Valores projetados de retornos de preços de contratos futuros da SALESFORCE", y="$US" ) +
  guides(colour = "none")

References

Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7(1), 77–91.
Link

Sharpe, W. F. (1966). Mutual Fund Performance. The Journal of Business, 39(1), 119–138.
Link

Elton, E. J., Gruber, M. J., Brown, S. J., & Goetzmann, W. N. (2007). Modern Portfolio Theory and Investment Analysis (9th ed.). Wiley.

Hilpisch, Y. (2018). Python for Finance: Mastering Data-Driven Finance. O’Reilly Media.

--- title: "Ciência de Dados para Negócios: Big Data for Finance Project" format: html: self-contained: true toc: true code-tools: true code-fold: true df-print: paged css: - styles.css - https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.1.1/css/all.min.css extensions: [fontawesome] editor: visual --- ------------------------------------------------------------------------ ::: callout-note ## Resumo <font size = 1> teste de futuro para as ações </font> ::: # Intro escrever ::: panel-tabset ## R ```{r setup, include=FALSE} knitr::opts_chunk$set( echo = TRUE, message = FALSE, warning = FALSE, comment = NA ) knitr::opts_chunk$set(comment = NA) # Remove all coments # of R outputs knitr::opts_chunk$set(warning = FALSE) # Remove all warnings # of R outputs knitr::opts_chunk$set(message = FALSE) # Remove all messages # of R outputs ``` ```{r} library(tidyverse) library(tidymodels) library(modeltime) library(timetk) library(purrr) library(tidyquant) library(tsibble) library(prophet) library(feasts) library(fable) library(fabletools) library(lubridate) library(tictoc) ``` Carregamos os dados: ```{r} tickers <- c( "BRFS3.SA", "JBSS3.SA", "BEEF3.SA", "MRFG3.SA", "TSN", "HRL", "GIS" ) ``` Então baixo os dados via Yahoo!Finance: ```{r} portfolioPrices <- NULL for ( Ticker in tickers ) portfolioPrices <- cbind( portfolioPrices, quantmod::getSymbols.yahoo( Ticker, from = "2019-01-01", auto.assign = FALSE )[,4] ) portfolioPrices <- portfolioPrices[apply(portfolioPrices, 1, function(x) all(!is.na(x))),] colnames(portfolioPrices) <- c( "BRFS3.SA", "JBSS3.SA", "BEEF3.SA", "MRFG3.SA", "TSN", "HRL", "GIS" ) # Visualizar com DT #DT::datatable(tail(portfolioPrices), options = list(pageLength = 10, scrollX = TRUE)) ``` Visualizando os dados dos nossos últimos retornos dos preços, temos: ```{r fig.width=9, fig.height=9} log_returns <- log(portfolioPrices) - log(lag(portfolioPrices)) log_returns <- na.omit(log_returns) log_returns <- log_returns |> timetk::tk_tbl(preserve_index = TRUE, rename_index = "date") tail(log_returns) ``` ```{r fig.width=9, fig.height=9} ln_returns <- log_returns ln_returns |> as.data.frame() |> dplyr::mutate( time = seq_along( TSN ) ) |> select(-date) |> tidyr::pivot_longer( !time, names_to = "Variables", values_to = "Value" ) |> dplyr::group_by(Variables) |> timetk::plot_time_series( time, Value, .interactive = F, # Change for TRUE for better visualization .facet_ncol = 2, .smooth = FALSE ) + ggplot2::theme( strip.background = ggplot2::element_rect(fill = "white", colour = "white") ) ``` ### Modelagem com fpp3 e validação cruzada temporal Precisaremos fazer um forecasting de curto prazo com nossos dados históricos de retornos pra formularmos nossas recomendações posteriores de compra, venda e espera: - Vamos começar com uma série por vez $\Rightarrow$ TSN ```{r} # Primeiro converto pra tsibble lnretTSN <- log_returns |> select(date, TSN) |> as_tsibble(index = date) glimpse(lnretTSN) ``` ```{r} treino <- lnretTSN |> filter_index(~"2025-01-01") ``` War models ```{r} tic() Modelos <- treino |> model( AjusteExp = ETS(TSN ~ error("A") + trend("N") + season("N")), # Ajuste Exponencial com auto AjExp_aditivo = ETS(TSN ~ error("A") + trend("A") + season("A")), # Ajuste Exponencial Aditivo AjExp_multiplicativo = ETS(TSN ~ error("M") + trend("A") + season("M")), # Ajuste Exponencial Multiplicativo Croston = CROSTON(TSN), # Modelo Croston HoltWinters = ETS(TSN ~ error("M") + trend("Ad") + season("M")), # Holt Winters Holt = ETS(TSN ~ error("A") + trend("A") + season("N")), # Holt HoltAmort = ETS(TSN ~ error("A") + trend("Ad", phi = 0.9) + season("N")), # Holt Amortecida Regr_Comp = TSLM(TSN ~ trend() + season()), # Regressao com tendencia e sazonalidade auto Regr_Harmonica = TSLM(TSN ~ trend() + fourier(K = 2)), # Regressao harmonica Regr_Quebras = TSLM(TSN ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais Snaive = SNAIVE(TSN), # SNAIVE Naive = NAIVE(TSN), #NAIVE Media_Movel = ARIMA(TSN ~ pdq(0,0,1)), # Media Movel Simples autoARIMA = ARIMA(TSN, stepwise = FALSE, approx = FALSE), # Auto ARIMA autoARIMA_saz = ARIMA(TSN, stepwise = FALSE, approx = FALSE, seasonal = TRUE), # AutoARIMA Sazonal # Regr_erros_ARIMA = auto.arima(TSN, xreg = fourier(K = 3), seasonal = FALSE), # Regressao com erros ARIMA ARIMA_saz_012011 = ARIMA(TSN ~ pdq(0,1,2) + PDQ(0,1,1)), # ARIMA Sazonal ordem 012011 ARIMA_saz_210011 = ARIMA(TSN ~ pdq(2,1,0) + PDQ(0,1,1)), # ARIMA Sazonal ordem 210011 ARIMA_saz_0301012 = ARIMA(TSN ~ 0 + pdq(3,0,1) + PDQ(0,1,2)), # ARIMA sazonal ARIMA_quad = ARIMA(TSN ~ I(trend()^2)), # ARIMA com tendencia temporal quadratica ARIMA_determ = ARIMA(TSN ~ 1 + trend() + pdq(d = 0)), # ARIMA com tendencia deterministica ARIMA_estocastico = ARIMA(TSN ~ pdq(d = 1)), # ARIMA com tendência estocastica Regr_Harm_dinamica = ARIMA(TSN ~ fourier(K=2) + PDQ(0,0,0)), # Regressao Harmonica Dinamica Regr_Harm_Din_MultSaz = ARIMA(TSN ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = 7*30, K = 10) + fourier(period = 7*30, K = 5)), Regr_Harm_Din_Saz = ARIMA(TSN ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = "month", K = 10) + fourier(period = "year", K = 2) ), # Rgr Harm Mult Saz Complexa # Auto_Prophet = prophet(TSN), # Auto prophet # Prophet_mult = prophet(TSN ~ season(period = "month", order = 2, type = "multiplicative")), # Prophet_aditivo = prophet(TSN ~ season(period = "month", order = 2, type = "additive")), # Prophet_geom = prophet(TSN ~ growth("geometric") + season(period = "month", order = 2, type = "multiplicative")), # Prophet_memo = prophet(TSN ~ growth("geometric") + season(period = "month", order = 5) + # season(period = "year", order = 2, type = "multiplicative")), Modelo_VAR = VAR(TSN, ic = "bic"), # Vetor Autoregressivo Random_Walk = RW(TSN ~ drift()), # Random Walk com drift Rede_Neural_AR = NNETAR(TSN, bootstrap = TRUE)#, # Rede Neural com auto AR e bootstraping nos erros # x11 = X_13ARIMA_SEATS(TSN ~ x11()) # X11 ARIMA Seats ) |> forecast(h = "24 months") # Horizonte de projecao para os proximos 30 dias apos corte no treino toc() ``` Selecionamos o melhor modelo (1 fold de validação cruzada somente): ```{r} Modelos |> accuracy(lnretTSN) |> arrange(RMSE) # Seleção da acuracia pelo menor RMSE para o conjunto de modelos ``` Gero um cenário com o modelo: ```{r} fit <- lnretTSN |> model( Regr_Quebras = TSLM(TSN ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais ) sim <- fit |> generate(h = 30, times = 5, bootstrap = TRUE) ``` Plotamos os forecasts com esse modelo pra três cenários distintos no futuro: ```{r fig.width=9, fig.height=4} lnretTSN |> filter_index("2025-01-01"~.) |> ggplot(aes(x = date)) + geom_line(aes(y = TSN)) + geom_line(aes(y = .sim, colour = as.factor(.rep)), data = sim) + labs(title="Valores projetados de retornos de preços de contratos futuros da TSN", y="$US" ) + guides(colour = "none") ``` ```{r} # Primeiro converto pra tsibble lnretGIS <- log_returns |> select(date, GIS) |> as_tsibble(index = date) glimpse(lnretGIS) ``` ```{r} treino <- lnretGIS |> filter_index(~"2025-01-01") ``` ```{r} tic() Modelos <- treino |> model( AjusteExp = ETS(GIS ~ error("A") + trend("N") + season("N")), # Ajuste Exponencial com auto AjExp_aditivo = ETS(GIS ~ error("A") + trend("A") + season("A")), # Ajuste Exponencial Aditivo AjExp_multiplicativo = ETS(GIS ~ error("M") + trend("A") + season("M")), # Ajuste Exponencial Multiplicativo Croston = CROSTON(GIS), # Modelo Croston HoltWinters = ETS(GIS ~ error("M") + trend("Ad") + season("M")), # Holt Winters Holt = ETS(GIS ~ error("A") + trend("A") + season("N")), # Holt HoltAmort = ETS(GIS ~ error("A") + trend("Ad", phi = 0.9) + season("N")), # Holt Amortecida Regr_Comp = TSLM(GIS ~ trend() + season()), # Regressao com tendencia e sazonalidade auto Regr_Harmonica = TSLM(GIS ~ trend() + fourier(K = 2)), # Regressao harmonica Regr_Quebras = TSLM(GIS ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais Snaive = SNAIVE(GIS), # SNAIVE Naive = NAIVE(GIS), #NAIVE Media_Movel = ARIMA(GIS ~ pdq(0,0,1)), # Media Movel Simples autoARIMA = ARIMA(GIS, stepwise = FALSE, approx = FALSE), # Auto ARIMA autoARIMA_saz = ARIMA(GIS, stepwise = FALSE, approx = FALSE, seasonal = TRUE), # AutoARIMA Sazonal # Regr_erros_ARIMA = auto.arima(TSN, xreg = fourier(K = 3), seasonal = FALSE), # Regressao com erros ARIMA ARIMA_saz_012011 = ARIMA(GIS ~ pdq(0,1,2) + PDQ(0,1,1)), # ARIMA Sazonal ordem 012011 ARIMA_saz_210011 = ARIMA(GIS ~ pdq(2,1,0) + PDQ(0,1,1)), # ARIMA Sazonal ordem 210011 ARIMA_saz_0301012 = ARIMA(GIS ~ 0 + pdq(3,0,1) + PDQ(0,1,2)), # ARIMA sazonal ARIMA_quad = ARIMA(GIS ~ I(trend()^2)), # ARIMA com tendencia temporal quadratica ARIMA_determ = ARIMA(GIS ~ 1 + trend() + pdq(d = 0)), # ARIMA com tendencia deterministica ARIMA_estocastico = ARIMA(GIS ~ pdq(d = 1)), # ARIMA com tendência estocastica Regr_Harm_dinamica = ARIMA(GIS ~ fourier(K=2) + PDQ(0,0,0)), # Regressao Harmonica Dinamica Regr_Harm_Din_MultSaz = ARIMA(GIS ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = 7*30, K = 10) + fourier(period = 7*30, K = 5)), Regr_Harm_Din_Saz = ARIMA(GIS ~ PDQ(0, 0, 0) + pdq(d = 0) + fourier(period = "month", K = 10) + fourier(period = "year", K = 2) ), # Rgr Harm Mult Saz Complexa # Auto_Prophet = prophet(TSN), # Auto prophet # Prophet_mult = prophet(TSN ~ season(period = "month", order = 2, type = "multiplicative")), # Prophet_aditivo = prophet(TSN ~ season(period = "month", order = 2, type = "additive")), # Prophet_geom = prophet(TSN ~ growth("geometric") + season(period = "month", order = 2, type = "multiplicative")), # Prophet_memo = prophet(TSN ~ growth("geometric") + season(period = "month", order = 5) + # season(period = "year", order = 2, type = "multiplicative")), Modelo_VAR = VAR(GIS, ic = "bic"), # Vetor Autoregressivo Random_Walk = RW(GIS ~ drift()), # Random Walk com drift Rede_Neural_AR = NNETAR(GIS, bootstrap = TRUE)#, # Rede Neural com auto AR e bootstraping nos erros # x11 = X_13ARIMA_SEATS(TSN ~ x11()) # X11 ARIMA Seats ) |> forecast(h = "24 months") # Horizonte de projecao para os proximos 30 dias apos corte no treino toc() ``` ```{r} fit <- lnretGIS |> model( Regr_Quebras = TSLM(GIS ~ trend(knots = c(2018, 2019, 2020))), # Regressao com quebras estruturais ) sim <- fit |> generate(h = 30, times = 5, bootstrap = TRUE) ``` ```{r fig.width=9, fig.height=4} lnretTSN |> filter_index("2025-01-01"~.) |> ggplot(aes(x = date)) + geom_line(aes(y = TSN)) + geom_line(aes(y = .sim, colour = as.factor(.rep)), data = sim) + labs(title="Valores projetados de retornos de preços de contratos futuros da SALESFORCE", y="$US" ) + guides(colour = "none") ``` ::: ------------------------------------------------------------------------ # References ------------------------------------------------------------------------ **Markowitz, H. (1952).** *Portfolio Selection*. The Journal of Finance, 7(1), 77–91.\ [Link](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1540-6261.1952.tb01525.x) **Sharpe, W. F. (1966).** *Mutual Fund Performance*. The Journal of Business, 39(1), 119–138.\ [Link](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1540-6261.1966.tb02865.x) **Elton, E. J., Gruber, M. J., Brown, S. J., & Goetzmann, W. N. (2007).** *Modern Portfolio Theory and Investment Analysis* (9th ed.). Wiley. **Hilpisch, Y. (2018).** *Python for Finance: Mastering Data-Driven Finance*. O'Reilly Media.